Text Generation
Transformers
Safetensors
mixtral
kv-cache-compression
inference-optimization
memory-efficient
custom_code
text-generation-inference
Instructions to use miike-ai/LeanMixtral-8x7B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use miike-ai/LeanMixtral-8x7B with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="miike-ai/LeanMixtral-8x7B", trust_remote_code=True)# Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("miike-ai/LeanMixtral-8x7B", trust_remote_code=True) model = AutoModelForCausalLM.from_pretrained("miike-ai/LeanMixtral-8x7B", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use miike-ai/LeanMixtral-8x7B with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "miike-ai/LeanMixtral-8x7B" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "miike-ai/LeanMixtral-8x7B", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/miike-ai/LeanMixtral-8x7B
- SGLang
How to use miike-ai/LeanMixtral-8x7B with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "miike-ai/LeanMixtral-8x7B" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "miike-ai/LeanMixtral-8x7B", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "miike-ai/LeanMixtral-8x7B" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "miike-ai/LeanMixtral-8x7B", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use miike-ai/LeanMixtral-8x7B with Docker Model Runner:
docker model run hf.co/miike-ai/LeanMixtral-8x7B
| { | |
| "architectures": [ | |
| "LeanMixtralForCausalLM" | |
| ], | |
| "attention_dropout": 0.0, | |
| "auto_map": { | |
| "AutoModelForCausalLM": "modeling_lean_mixtral.LeanMixtralForCausalLM" | |
| }, | |
| "bos_token_id": 1, | |
| "dtype": "bfloat16", | |
| "eos_token_id": 2, | |
| "head_dim": null, | |
| "hidden_act": "silu", | |
| "hidden_size": 4096, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 14336, | |
| "leanmix_compressed_kv_layers": [ | |
| 3, | |
| 8, | |
| 13, | |
| 14, | |
| 16, | |
| 22, | |
| 23, | |
| 24, | |
| 25, | |
| 26, | |
| 27, | |
| 28 | |
| ], | |
| "leanmix_kv_decoder_depth": { | |
| "13": 2, | |
| "14": 2, | |
| "16": 2, | |
| "22": 2, | |
| "23": 2, | |
| "24": 2, | |
| "25": 2, | |
| "26": 2, | |
| "27": 2, | |
| "28": 2, | |
| "3": 2, | |
| "8": 2 | |
| }, | |
| "leanmix_kv_decoder_hidden_dim": { | |
| "13": 128, | |
| "14": 128, | |
| "16": 128, | |
| "22": 128, | |
| "23": 128, | |
| "24": 128, | |
| "25": 128, | |
| "26": 128, | |
| "27": 128, | |
| "28": 128, | |
| "3": 128, | |
| "8": 128 | |
| }, | |
| "leanmix_kv_fixed_alpha": { | |
| "13": 1.0, | |
| "14": 1.0, | |
| "16": 1.0, | |
| "22": 1.0, | |
| "23": 1.0, | |
| "24": 1.0, | |
| "25": 1.0, | |
| "26": 1.0, | |
| "27": 1.0, | |
| "28": 1.0, | |
| "3": 1.0, | |
| "8": 1.0 | |
| }, | |
| "leanmix_kv_input_dims": { | |
| "13": 1024, | |
| "14": 1024, | |
| "16": 1024, | |
| "22": 1024, | |
| "23": 1024, | |
| "24": 1024, | |
| "25": 1024, | |
| "26": 1024, | |
| "27": 1024, | |
| "28": 1024, | |
| "3": 1024, | |
| "8": 1024 | |
| }, | |
| "leanmix_kv_key_dims": { | |
| "13": 4, | |
| "14": 4, | |
| "16": 4, | |
| "22": 4, | |
| "23": 4, | |
| "24": 4, | |
| "25": 4, | |
| "26": 4, | |
| "27": 4, | |
| "28": 4, | |
| "3": 4, | |
| "8": 4 | |
| }, | |
| "leanmix_kv_residual_decoder": { | |
| "13": false, | |
| "14": false, | |
| "16": false, | |
| "22": false, | |
| "23": false, | |
| "24": false, | |
| "25": false, | |
| "26": false, | |
| "27": false, | |
| "28": false, | |
| "3": false, | |
| "8": false | |
| }, | |
| "leanmix_kv_value_dims": { | |
| "13": 4, | |
| "14": 4, | |
| "16": 4, | |
| "22": 4, | |
| "23": 4, | |
| "24": 4, | |
| "25": 4, | |
| "26": 4, | |
| "27": 4, | |
| "28": 4, | |
| "3": 4, | |
| "8": 4 | |
| }, | |
| "max_position_embeddings": 32768, | |
| "model_type": "mixtral", | |
| "num_attention_heads": 32, | |
| "num_experts_per_tok": 2, | |
| "num_hidden_layers": 32, | |
| "num_key_value_heads": 8, | |
| "num_local_experts": 8, | |
| "output_router_logits": false, | |
| "rms_norm_eps": 1e-05, | |
| "rope_theta": 1000000.0, | |
| "router_aux_loss_coef": 0.02, | |
| "router_jitter_noise": 0.0, | |
| "sliding_window": null, | |
| "tie_word_embeddings": false, | |
| "transformers_version": "4.57.6", | |
| "use_cache": true, | |
| "vocab_size": 32000 | |
| } | |